Cancers occur naturally so you need an exposure model to give you an idea of what to expect, and then compare that to the background of normal cancer. Even assuming that he checked mortality records which I assume he must have, he still had to apply statistical filter to make it meaningful.

Interesting. My view as a dilettante on the sidelines was that LNT was on it's way to being discredited. i.e. along the lines of the last paragraph here: https://rationalwiki.org/wiki/Linear_no-threshold

I'd be interested to know more about the current balance of thinking on the topic. The optimist in me obviously would prefer a less awful relationship between exposure and mortality but sadly I can't impose my wishes on reality.